solodot-mcp
MCP server for solo business founders to route painpoints, generate execution assets, and maintain control through approval workflow.
README
solodot-mcp
connect the dots that scale your solo business.
mcp server for solodot. route the active founder painpoint, generate one execution asset, and keep the founder in control through approve, revise, or reject.
features
- six execution packs: bottleneck, customers, founder functions, learning, trust, and agent workflows.
- five tools: route, generate, review, inspect, and export.
- local stdio for codex and claude desktop.
- authenticated streamable http for remote agents.
- anthropic and vertex model providers.
- in-memory local runs or durable supabase runs.
- docker and apple container support.
- no silent email, slack, crm, publishing, invoice, or external actions.
requirements
- node.js 20+
- an anthropic or vertex api key
- supabase for remote http mode
- codex, claude, or another mcp client
apple container requires apple silicon and macos 26. install it from the official releases, then start the service with container system start.
install
cd /Users/0x79de/dev/solodot-mcp
npm install
cp .env.example .env
npm run build
set at least one provider in .env:
SOLODOT_DEFAULT_PROVIDER=anthropic
ANTHROPIC_API_KEY=...
use SOLODOT_DEFAULT_PROVIDER=vertex with VERTEX_API_KEY for vertex.
setup
codex
add a local stdio server to ~/.codex/config.toml:
[mcp_servers.solodot]
command = "node"
args = ["/Users/0x79de/dev/solodot-mcp/dist/stdio.js"]
env = { SOLODOT_DEFAULT_PROVIDER = "anthropic", ANTHROPIC_API_KEY = "your-key" }
claude desktop
add this to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"solodot": {
"command": "node",
"args": ["/Users/0x79de/dev/solodot-mcp/dist/stdio.js"],
"env": {
"SOLODOT_DEFAULT_PROVIDER": "anthropic",
"ANTHROPIC_API_KEY": "your-key"
}
}
}
}
restart the client after changing its config. ready-to-copy examples live in examples/.
usage
after setup, ask your agent:
- "route the biggest painpoint in my solo business."
- "turn this route into the execution asset i should use this week."
- "revise the asset for a narrower b2b saas buyer."
- "approve it and export the run as markdown."
the normal flow is:
route_painpoint
-> generate_execution_asset
-> review_execution_asset
-> export_solodot_run
approval records founder intent. it never executes the asset outside solodot.
remote http
remote mode serves POST /mcp and GET /health on port 8787.
configure:
SUPABASE_URL=https://your-project.supabase.co
SUPABASE_SECRET_KEY=...
SOLODOT_MCP_API_KEYS=long-random-founder-token
ANTHROPIC_API_KEY=...
run:
npm run build
npm run start:http
remote codex config:
[mcp_servers.solodot]
url = "https://mcp.example.com/mcp"
bearer_token_env_var = "SOLODOT_MCP_TOKEN"
use the same public https endpoint and bearer token for a claude custom connector or another remote mcp host.
apple container
solodot uses the same oci image for apple container and docker.
container system start
npm run container:build
npm run container:run
container:run reads .env, starts solodot-mcp with an init process, and publishes only 127.0.0.1:8787.
npm run container:health
npm run container:logs
npm run container:stop
override defaults when needed:
SOLODOT_MCP_ENV_FILE=/absolute/path/to/solodot.env \
SOLODOT_MCP_PORT=9000 \
npm run container:run
apple container runs linux containers as lightweight virtual machines on apple silicon. the image remains portable to any oci-compatible runtime.
docker
build from the standalone mcp repository:
cd /Users/0x79de/dev/solodot-mcp
docker build -t solodot-mcp:local .
docker run --init --rm --env-file .env -p 127.0.0.1:8787:8787 solodot-mcp:local
tools
route_painpoint: choose exactly one execution pack.generate_execution_asset: create the full asset for that fixed route.review_execution_asset: approve, revise, or reject.get_solodot_run: inspect diagnosis, assets, and approvals.export_solodot_run: return an approved asset as markdown or json.
resources are available at solodot://execution-packs, solodot://approval-boundaries, and solodot://product-context.
slack
an mcp-capable agent can call solodot while its conversation lives in slack.
this is not a native slack bot. it does not read channels, receive slash commands, or post messages.
development
npm run typecheck
npm test
npm run build
inspect the local server:
npx @modelcontextprotocol/inspector node dist/stdio.js
links
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。